Cardiac anomaly detection based on time and frequency domain features using tree-based classifiers

Martin Kropf, D. Hayn, D. Morris, Aravind Kumar Radhakrishnan, E. Belyavskiy, A. Frydas, E. Pieske-Kraigher, B. Pieske, G. Schreier

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Objective: Recent advantages in mHealth-enabled ECG recorders boosted the demand for algorithms, which are able to automatically detect cardiac anomalies with high accuracy. Approach: We present a combined method of classical signal analysis and machine learning which has been developed during the Computing in Cardiology Challenge (CinC) 2017. Almost 400 hand-crafted features have been developed to reflect the complex physiology of cardiac arrhythmias and their appearance in single-channel ECG recordings. For the scope of this article, we performed several experiments on the publicly available challenge dataset to improve the classification accuracy. We compared the performance of two tree-based algorithms - gradient boosted trees and random forests - using different parameters for learning. We assessed the influence of five different sets of training annotations on the classifiers performance. Further, we present a new web-based ECG viewer to review and correct the training labels of a signal data set. Moreover, we analysed the feature importance and evaluated the model performance when using only a subset of the features. The primary data source used in the analysis was the dataset of the CinC 2017, consisting of 8528 signals from four classes. Our best results were achieved using a gradient boosted tree model which worked significantly better than random forests. Main results: Official results of the challenge follow-up phase provided by the Challenge organizers on the full hidden test set are 90.8% (Normal), 84.1% (AF), 74.5% (Other), resulting in a mean F1-score of 83.2%, which was only 1.6% behind the challenge winner and 0.2% ahead of the next-best algorithm. Official results were rounded to two decimal places which lead to the equal-second best F1F-score of 83% with five others. Significance: The algorithm achieved the second-best score among 80 algorithms of the Challenge follow-up phase equal with five others.

Original languageEnglish
Article number114001
JournalPhysiological measurement
Volume39
Issue number11
DOIs
Publication statusPublished - 25 Oct 2018

Fingerprint

Classifiers
Electrocardiography
Cardiology
Telemedicine
Information Storage and Retrieval
Signal analysis
Physiology
Learning systems
Labels
Cardiac Arrhythmias
Hand
Learning
Datasets
Experiments
Forests

Keywords

  • cardiac arrhythmias
  • ECG classification
  • machine learning
  • predictive modelling

ASJC Scopus subject areas

  • Biophysics
  • Physiology
  • Biomedical Engineering
  • Physiology (medical)

Cite this

Cardiac anomaly detection based on time and frequency domain features using tree-based classifiers. / Kropf, Martin; Hayn, D.; Morris, D.; Radhakrishnan, Aravind Kumar; Belyavskiy, E.; Frydas, A.; Pieske-Kraigher, E.; Pieske, B.; Schreier, G.

In: Physiological measurement, Vol. 39, No. 11, 114001, 25.10.2018.

Research output: Contribution to journalArticleResearchpeer-review

Kropf, M, Hayn, D, Morris, D, Radhakrishnan, AK, Belyavskiy, E, Frydas, A, Pieske-Kraigher, E, Pieske, B & Schreier, G 2018, 'Cardiac anomaly detection based on time and frequency domain features using tree-based classifiers' Physiological measurement, vol. 39, no. 11, 114001. https://doi.org/10.1088/1361-6579/aae13e
Kropf, Martin ; Hayn, D. ; Morris, D. ; Radhakrishnan, Aravind Kumar ; Belyavskiy, E. ; Frydas, A. ; Pieske-Kraigher, E. ; Pieske, B. ; Schreier, G. / Cardiac anomaly detection based on time and frequency domain features using tree-based classifiers. In: Physiological measurement. 2018 ; Vol. 39, No. 11.
@article{cecc2113385343d990099a1df4add6b1,
title = "Cardiac anomaly detection based on time and frequency domain features using tree-based classifiers",
abstract = "Objective: Recent advantages in mHealth-enabled ECG recorders boosted the demand for algorithms, which are able to automatically detect cardiac anomalies with high accuracy. Approach: We present a combined method of classical signal analysis and machine learning which has been developed during the Computing in Cardiology Challenge (CinC) 2017. Almost 400 hand-crafted features have been developed to reflect the complex physiology of cardiac arrhythmias and their appearance in single-channel ECG recordings. For the scope of this article, we performed several experiments on the publicly available challenge dataset to improve the classification accuracy. We compared the performance of two tree-based algorithms - gradient boosted trees and random forests - using different parameters for learning. We assessed the influence of five different sets of training annotations on the classifiers performance. Further, we present a new web-based ECG viewer to review and correct the training labels of a signal data set. Moreover, we analysed the feature importance and evaluated the model performance when using only a subset of the features. The primary data source used in the analysis was the dataset of the CinC 2017, consisting of 8528 signals from four classes. Our best results were achieved using a gradient boosted tree model which worked significantly better than random forests. Main results: Official results of the challenge follow-up phase provided by the Challenge organizers on the full hidden test set are 90.8{\%} (Normal), 84.1{\%} (AF), 74.5{\%} (Other), resulting in a mean F1-score of 83.2{\%}, which was only 1.6{\%} behind the challenge winner and 0.2{\%} ahead of the next-best algorithm. Official results were rounded to two decimal places which lead to the equal-second best F1F-score of 83{\%} with five others. Significance: The algorithm achieved the second-best score among 80 algorithms of the Challenge follow-up phase equal with five others.",
keywords = "cardiac arrhythmias, ECG classification, machine learning, predictive modelling",
author = "Martin Kropf and D. Hayn and D. Morris and Radhakrishnan, {Aravind Kumar} and E. Belyavskiy and A. Frydas and E. Pieske-Kraigher and B. Pieske and G. Schreier",
year = "2018",
month = "10",
day = "25",
doi = "10.1088/1361-6579/aae13e",
language = "English",
volume = "39",
journal = "Physiological measurement",
issn = "0967-3334",
publisher = "IOP Publishing Ltd.",
number = "11",

}

TY - JOUR

T1 - Cardiac anomaly detection based on time and frequency domain features using tree-based classifiers

AU - Kropf, Martin

AU - Hayn, D.

AU - Morris, D.

AU - Radhakrishnan, Aravind Kumar

AU - Belyavskiy, E.

AU - Frydas, A.

AU - Pieske-Kraigher, E.

AU - Pieske, B.

AU - Schreier, G.

PY - 2018/10/25

Y1 - 2018/10/25

N2 - Objective: Recent advantages in mHealth-enabled ECG recorders boosted the demand for algorithms, which are able to automatically detect cardiac anomalies with high accuracy. Approach: We present a combined method of classical signal analysis and machine learning which has been developed during the Computing in Cardiology Challenge (CinC) 2017. Almost 400 hand-crafted features have been developed to reflect the complex physiology of cardiac arrhythmias and their appearance in single-channel ECG recordings. For the scope of this article, we performed several experiments on the publicly available challenge dataset to improve the classification accuracy. We compared the performance of two tree-based algorithms - gradient boosted trees and random forests - using different parameters for learning. We assessed the influence of five different sets of training annotations on the classifiers performance. Further, we present a new web-based ECG viewer to review and correct the training labels of a signal data set. Moreover, we analysed the feature importance and evaluated the model performance when using only a subset of the features. The primary data source used in the analysis was the dataset of the CinC 2017, consisting of 8528 signals from four classes. Our best results were achieved using a gradient boosted tree model which worked significantly better than random forests. Main results: Official results of the challenge follow-up phase provided by the Challenge organizers on the full hidden test set are 90.8% (Normal), 84.1% (AF), 74.5% (Other), resulting in a mean F1-score of 83.2%, which was only 1.6% behind the challenge winner and 0.2% ahead of the next-best algorithm. Official results were rounded to two decimal places which lead to the equal-second best F1F-score of 83% with five others. Significance: The algorithm achieved the second-best score among 80 algorithms of the Challenge follow-up phase equal with five others.

AB - Objective: Recent advantages in mHealth-enabled ECG recorders boosted the demand for algorithms, which are able to automatically detect cardiac anomalies with high accuracy. Approach: We present a combined method of classical signal analysis and machine learning which has been developed during the Computing in Cardiology Challenge (CinC) 2017. Almost 400 hand-crafted features have been developed to reflect the complex physiology of cardiac arrhythmias and their appearance in single-channel ECG recordings. For the scope of this article, we performed several experiments on the publicly available challenge dataset to improve the classification accuracy. We compared the performance of two tree-based algorithms - gradient boosted trees and random forests - using different parameters for learning. We assessed the influence of five different sets of training annotations on the classifiers performance. Further, we present a new web-based ECG viewer to review and correct the training labels of a signal data set. Moreover, we analysed the feature importance and evaluated the model performance when using only a subset of the features. The primary data source used in the analysis was the dataset of the CinC 2017, consisting of 8528 signals from four classes. Our best results were achieved using a gradient boosted tree model which worked significantly better than random forests. Main results: Official results of the challenge follow-up phase provided by the Challenge organizers on the full hidden test set are 90.8% (Normal), 84.1% (AF), 74.5% (Other), resulting in a mean F1-score of 83.2%, which was only 1.6% behind the challenge winner and 0.2% ahead of the next-best algorithm. Official results were rounded to two decimal places which lead to the equal-second best F1F-score of 83% with five others. Significance: The algorithm achieved the second-best score among 80 algorithms of the Challenge follow-up phase equal with five others.

KW - cardiac arrhythmias

KW - ECG classification

KW - machine learning

KW - predictive modelling

UR - http://www.scopus.com/inward/record.url?scp=85055658886&partnerID=8YFLogxK

U2 - 10.1088/1361-6579/aae13e

DO - 10.1088/1361-6579/aae13e

M3 - Article

VL - 39

JO - Physiological measurement

JF - Physiological measurement

SN - 0967-3334

IS - 11

M1 - 114001

ER -